In this office hours call, I did two things live: first, I tested Replit AI’s brand-new mobile app feature (it had literally just come out a few days earlier), and then I opened it up for student Q&A. If you’ve been following my earlier experiments, you know I tried building a mobile app with a different tool in late summer/early fall—and it was clunky enough that I stopped. This time, I wanted to see if Replit could actually get us to a working iOS-style app experience without me writing code. I started with a super simple two-screen app prompt just to validate the workflow, and Replit immediately generated branding, copy, and navigation. Then I pushed it further: I asked how to pull “fastest growing cryptocurrencies” over different timeframes. CoinGecko sounded perfect on paper, but the free tier wasn’t available the way the AI suggested, so I switched to Dexscreener (no API key). Replit integrated the API, but we hit typical AI-builder issues—duplicate tokens, weird results, and a not-so-smart query approach—so I iterated and had it fix the logic. The big takeaway: this is a fast way to prototype an app that would normally cost real money and time with an engineer, but you still have to expect bugs and debugging. Replit is great for simple builds and front-end; when business logic gets complex, I often cross-check with Claude (and sometimes ChatGPT) so the AIs “check each other” and I do less thinking.